Effectiveness of the public health policy for breast cancer screening in Finland: population based cohort studyBMJ 1997; 314 doi: http://dx.doi.org/10.1136/bmj.314.7084.864 (Published 22 March 1997) Cite this as: BMJ 1997;314:864
- Matti Hakama, professor of epidemiologya,
- Eero Pukkala, research scientistb,
- Minna HeikkilÄ, research assistantb,
- Mervi Kallio, research scientistb
- a University of Tampere School of Public Health Box 607 FIN-33101 Tampere Finland
- b Finnish Cancer Registry Liisankatu 21B FIN-00170 Helsinki
- Correspondence to: Dr Hakama
- Accepted 16 January 1997
Objective: To evaluate the effectiveness of screening for breast cancer as a public health policy.
Design: Follow up in 1987-92 of Finnish women invited to join the screening programme in 1987-9 and of the control women (balanced by age and matched by municipality of residence), who were not invited to the service screening.
Subjects: Of the Finnish women born in 1927-39, 89 893 women invited for screening and 68 862 controls were followed; 1584 breast cancers were diagnosed.
Main outcome measures: Rate ratio of deaths from breast cancer among the women invited for screening to deaths among those not invited.
Results: There were 385 deaths from breast cancer, of which 127 were among the 1584 incident cases in 1987-92. The rate ratio of death was 0.76 (95% confidence interval 0.53 to 1.09). The effect was larger and significant (0.56; 0.33 to 0.95) among women aged under 56 years at entry. 20 cancers were prevented (one death prevented per 10 000 screens).
Conclusions: A breast screening programme can achieve a similar effect on mortality as achieved by the trials for breast cancer screening. However, it may be difficult to justify a screening programme as a public health policy on the basis of the mortality reduction only. Whether to run a screening programme as a public health policy also depends on its effects on the quality of life of the target population and what the resources would be used for if screening was not done. Given all the different dimensions in the effect, mammography based breast screening is probably justifiable as a public health policy.
Several countries have a breast cancer screening programme, but none has yet reported this as resulting in a reduction of breast cancer mortality
This study shows that a breast cancer screening programme can achieve similar reduction in mortality to that seen in randomised trials
Effects on quality of life, cost of breast cancer screening, and the alternative use of resources should affect the decision whether to introduce a screening programme
The first randomised, population based trial on breast cancer screening based on mammography was in the 1960s.1 This study showed that about one in three deaths from breast cancer can be prevented if women are screened. Later, similar results from Sweden2 3 and the Netherlands4 5 were published. Mammography based screening became widespread, and in several countries it has been part of a public health policy or organised screening programme. Nowhere has a screening programme been reported to result in reduction of breast cancer mortality–a reduction in mortality is the goal of any cancer screening programme. Finland was the first country with a nationwide screening programme. We report here its effectiveness.
Subjects and methods
In Finland, nationwide population based breast cancer screening was introduced in 1987. Women in birth cohorts recommended by the National Board of Health are individually identified and invited for screening. The programme covers women aged 50-59 years and can be continued up to age 64. Women are screened every two years.
The Cancer Society of Finland (and its regional member societies) established 11 mammography screening centres. Local municipalities (which in Finland are responsible for the public health services) were entitled to make an agreement with one of these screening centres. The programme organised by the cancer society covered two thirds of the 460 Finnish municipalities during the first years of the nationwide screening programme. In 1987, 84% of the municipalities with agreements with the cancer society followed the guidelines of the National Board of Health. The programme was introduced gradually with cohorts born in even years; the women born in odd years were controls during the first years of the programme (fig 1).
Each woman in the cohorts selected by the municipal council receives a letter with a personal appointment as well as details of the screening procedure. Every participant receives a letter notifying her whether the screen was positive or negative. The women with a positive result are given an appointment time for confirmation of the diagnosis. No reminders are sent to women who do not attend for screening. Two view mammography is used. Two radiologists interpret the mammograms, of whom one carries out further examinations in women with positive screen results.
A centralised mass screening registry for identification, invitation, and follow up of the women is part of the Finnish Cancer Registry, which operates nationwide and is population based. The National Population Registry, national registration of deaths, and cancer registrations are linked with the screening results by the mass screening registry.
We studied women born in 1927-39 living in the municipalities who were screened by the Cancer Society of Finland. We classified women invited in 1987-9 as either “screened” or “not screened” (those who did not attend for screening). The controls were women in the same municipalities as those screened, born in 1927, 1929, 1933, 1935, or 1939 (fig 1). The women born in 1931 and 1937 were recommended to be screened for the first time in 1989. As they potentially provided few person years and with short follow up, we excluded them from the present analyses. We also excluded women born in 1936 to achieve a balance in age between cases and controls. We identified, by linkage to the Finnish Cancer Registry, breast cancer cases diagnosed at screening, interval cancers, cancers diagnosed among the women invited but not screened, cancers diagnosed in the control cohorts, and deaths from breast cancer. The follow up was extended to the end of 1992.
We evaluated the effect of screening in terms of standardised mortality ratios among the women invited for screening compared with those in the controls, and we called the ratio of these two ratios the rate ratio. The comparison rates for the standardised mortality ratios were those for the whole of Finland during the total period of follow up. The mortality due to breast cancer was estimated by including and by excluding the cases of breast cancer diagnosed before the first screening round (“total” mortality and “refined” mortality respectively). The person years were estimated from the month of screening for those screened. For the controls and for women invited but not screened, the start of follow up was defined as the mean date of screening in the municipality in that year. The end of follow up was 31 December 1992, date of death, or date of migration to a foreign country, whichever occurred first. If a control was invited in 1987-9 to be screened she was moved from the control group to the group of invited women, and the woman years were distributed to the control group or the “invited” group according to the time of screening. Such invitations took place only if the woman's municipality of residence did not comply with the general recommendations.
Table 1) shows the numbers of women invited and screened and the controls by year of birth. Table 2) shows the numbers of new cases of breast cancer and deaths from breast cancer during 1987-92. Table 3) shows the standardised mortality ratios.
There were 64 deaths among the women invited for screening and 63 deaths among the controls from breast cancers diagnosed after the start of follow up (table 2). The standardised mortality ratio of refined mortality was higher among the women invited but not screened than among the controls (rate ratio=1.42). For the women screened the rate ratio was 0.67. This resulted in a total rate ratio of 0.76 (95% confidence interval 0.53 to 1.09) for the women invited for screening, which showed a 24% protective effect due to screening, which was not significant. The protective effect varied by the year of follow up; the effect emerged only during the three to four years of follow up (rate ratio=0.69) and was significant (0.35 to 0.99). Because this effect occurred relatively early, it was seen only for deaths that occurred before the age of 60 years. Therefore, the protective effect also differed by age at entry to the study. Among those born in 1927-30–that is, women mainly aged over 57 years at the time of the first screen–the effect was negligible. Those born in 1932 and later (mainly aged under 56 at entry) had a rate ratio for death from breast cancer of 0.56, which was significant (0.33 to 0.95) (table 4)4).
Had the refined standardised mortality ratio among women invited for screening been the same as in the control population, there would have been 84 deaths from breast cancer (64/0.76) among the invited women. As there were only 64 deaths, the number of deaths prevented because of the screening can be estimated to be 20.
Randomised trials show that screening with mammography reduces mortality from breast cancer, with an average reduction of about 30%.6 7 8 On the basis of such results, screening was introduced as a national public health policy or with an organised programme in several countries, including the United Kingdom,9 10 Sweden,11 the Netherlands,12 and Finland.13 Spontaneous or opportunistic screening is a widespread practice in several other countries. The effectiveness of such a public health policy in terms of a reduction in mortality was evaluated in the United Kingdom on the basis of the national rates before and after the introduction of screening.10 The evaluation found a decrease in mortality, which was, however, unlikely to be due to screening. Such a design may not be able to disclose the potential effect of screening because of small and gradual effect.14
Finland was the first country to introduce nationwide breast cancer screening as a public health policy. The participation rate in the first year, 1987, was 88%,13 which is among the highest rates reported anywhere, and the programme was successfully carried out. The programme first covered women born in even-year birth cohorts. The availability as controls of women born in odd years adjacent to the screened cohorts decreased during the four year implementation period as the programme expanded. In Finland the general health services are funded by the municipalities, who receive state subsidies for such purposes. Success of the design depended on the motivation of the municipalities to comply with the National Board of Health's guidelines on screening.
The potential bias due to self selection of municipalities was eliminated by having the controls from the same municipality as the women invited for screening. The analysis was based on 10 birth cohorts of women born 1927-39. All the control women belonged to five birth cohorts, two of which (1935, 1939) were recommended to be screened for the first time in 1990, two (1929, 1933) in 1991, and one (1927) not at all. It was assumed that this late screening would not substantially affect the deaths from breast cancer by the end of 1992. The controls were unbiased for age, because the trend in mortality from breast cancer is linear over the ages 50 years to 65 years, and the one year differences in women invited for screening and the controls were balanced.
Some of the municipalities began to organise screening for women born in these five control cohorts; the women who were invited for screening were removed from the group of controls at the time of first screening and were further classified according to their actual participation to prevent any effect due to dilution. Such changes were relatively few for women born in 1927, 1929, and 1933, but more than 30% of the women in cohorts born in 1935 and 1939 had to be removed from the control group.
We eliminated the obvious bias due to self selection (women attending for screening and women not attending) by comparing the mortality among all the women who were invited for screening (regardless of whether they attended) to that of the controls. We evaluated the basic risk of death by comparing the numbers of deaths from breast cancer among the women invited for screening and among the controls to that expected for the overall mortality for Finland. The total standardised mortality ratio for women invited but not screened was high, because breast cancer reduces the feasibility of attending the programme. The women invited for screening had a mortality from breast cancer equal to that expected on the basis of the total Finnish rates (standardised mortality ratio=1.0). This would point to ineffectiveness of the screening programme. However, the controls had a higher risk of death than expected (1.1), which indicates that municipalities in which women had a high risk of breast cancer were more likely to be included in our material. Because risk of breast cancer is high among more wealthy women, the more wealthy municipalities were more ready to start screening with the Cancer Society of Finland and its regional member societies.
There was only a narrow window by year of follow up to evaluate the Finnish programme. The effect of screening on mortality did not appear until the third and fourth year of follow up and then was lost because the controls were also gradually being screened. The difference in calendar years between the women screened first and those screened last was only four years, in line with the national recommendation. Therefore, because of dilution of screening in the controls, we could evaluate only the early effect. Several trials examining the effect of screening on mortality have shown a significant difference in cumulative deaths from breast cancer only a few years after the first round. However, the point estimates have been consistent, with a constant proportion of deaths prevented almost immediately after the first round of screening among women aged 50-65 years at entry.3 7 15 The delayed effect was a consistent finding only among those aged under 50 years at entry.3 7 15 In our study the effect in mortality rapidly disappeared as controls merged into the national programme. The disappearance of the effect is different from the experience in several randomised trials,1 3 7 where the difference of cumulative rates increased for many years after the programme ended or after the control arm was merged into a programme identical to that offered for the screening arm. For evaluation a period of implementation of the public health policy in Finland longer than four years would have been feasible. However, Finland had the resources to expand the programme relatively rapidly, and it was not justifiable to withhold the public health policy from more women than necessary.
An effective policy
During the follow up of 700 000 person years, 1600 breast cancers were diagnosed and 400 deaths from breast cancer occurred. Our final estimate of effect–a 24% reduction in mortality from breast cancer–was based on 127 refined deaths from breast cancer only. The reduction is not significant, but it is consistent with the results from randomised trials.2 3 8 It was larger (42%) and significant in women aged under 56 years at entry. Was the effect large enough to justify the time and resources spent on this public health policy? The small effect of breast cancer screening had already been pointed out 10 years previously,16 and scepticism is getting more common.17 In our study about 200 000 screening tests prevented 20 deaths. This result is similar to that found by Wright and Mueller.17
The screening programme also has effects other than reduction in deaths from breast cancer–namely, longer survival. Most women in Finland attend cancer screening programmes for reassurance that they do not have preclinical cancer.18 19 The quality of life of the patient is likely to improve–for example, through breast conserving surgery. If no screening programme operates the resources will go elsewhere. Breast cancer screening is cost effective compared with many other healthcare services.20 Much of the health service resources are spent with poor control of effectiveness.
We are indebted to Ms Pirkko Pakarinen and Ms Anita Pirinen from the Finnish Cancer Registry.
Funding Cancer Society of Finland, the Finnish Slot Machine Association, and the Finnish Cancer Institute.
Conflict of interest None.